目的对硫酸沙丁胺醇渗透泵控释片剂释药行为进行遗传神经网络预测。方法将人工神经网络与遗传算法应用于硫酸沙丁胺醇渗透泵控释片剂释药行为预测研究,提出了采用遗传算法对人工神经网络进行优化的网络模型建立方法,并以硫酸沙丁胺醇渗透泵控释片剂处方样本的相关实验数据为考察对象,考虑包衣液中的聚乙二醇1500含量(η)包衣膜厚度(δ) 对相关系数(r)各处方8h的累积释放度(F₈)影响,建立了硫酸沙丁胺醇渗透泵控释片剂释药行为遗传神经网络预测模型。结果通过实验测量与GA-BP神经网络预测结果比较,验证了经遗传算法优化的GA-BP神经网络模型对r的预测精度达 97.23％,对F8的预测精度为94.68％。结论由于该模型应用遗传算法对BP神经网络权值和学习进行优化,从而克服了BP神经网络训练速度慢,容易陷入局域极小和全局搜索能力弱等弱点,可以用于硫酸沙丁胺醇渗透泵控释片剂释药行为预测。

OBJECTIVE To pridict salbutamol sulfate osmotic pump tablet drug release by BP-GA neural network. A genetic-algorithm-based system using Artificial Neural Network for pridicting salbutamol sulfale osmotic pump tablet drug release. METHODS To realize the algorithm, a BP-GA neural network model was established.The data from the salbutamol sulfate osmotic pump tablet were analyzed,the effects of PEG1500 content η,the ratio of thickness δ to drug release coefficient γ and the 8 h drug release scale F₈ into were investigated. RESULTS Comparing the experiment results with that of simulations and analysis based on the BP-GA neural network, the precision of drug release coefficient γ and the precision of 8 h drug release scale F₈ were at 97.23% and 94.68% , respectively. CONCLUSION This model is based on the GA optimized the learning and weight of BP-NN. With the BP-GA neural network system, the weakness, such as the slow training speed of BP-NN and the vulnerability to local area pole smallness, are overcomed.It can be used in the prediction of salbutamol sulfate osmotic pump tablet drug release.